LocallyLinearEmbedding<T>

Models & Types · AiDotNet.Preprocessing.DimensionalityReduction

Locally Linear Embedding for nonlinear dimensionality reduction.

Inheritance: TransformerBase<T, Matrix<T>, Matrix<T>> → LocallyLinearEmbedding<T>

For Beginners

LLE preserves local relationships:

  • Each point is described by its neighbors
  • The weights describe “how much” each neighbor contributes
  • The embedding keeps these relationships intact
  • Good for unfolding curved manifolds (like the Swiss roll)

How It Works

LLE preserves local neighborhood structure by representing each point as a weighted linear combination of its neighbors. The embedding is found by preserving these reconstruction weights in lower dimensions.

The algorithm:

  1. Find k nearest neighbors for each point
  2. Compute reconstruction weights that best reconstruct each point from neighbors
  3. Find low-dimensional embedding that preserves these weights

Constructors

Constructor Summary
LocallyLinearEmbedding(int nComponents, int nNeighbors, double reg, LLEMethod method, int[] columnIndices) Creates a new instance of LocallyLinearEmbedding.

Properties

Property Summary
LLEMethod Method { get; } Gets the LLE method.
Matrix<T> Embedding { get; } Gets the embedding result.
bool SupportsInverseTransform { get; } Gets whether this transformer supports inverse transformation.
double Regularization { get; } Gets the regularization parameter.
int NComponents { get; } Gets the number of components.
int NNeighbors { get; } Gets the number of neighbors.

Methods

Method Summary
string[] GetFeatureNamesOut(string[] inputFeatureNames) Gets the output feature names after transformation.